top of page

The Retire Advocate 

May

2026

Nuclear Power for AI?

Peter Harris

Demand for new nuclear power is increasing, as shown in this year’s Washington state legislative session. House Bill 2090, which failed, would have “[integrated] advanced nuclear energy into the state energy strategy,” and would have let private parties – that is, industries desiring nuclear power – fund cost-benefit analysis and make recommendations. House Bill 1210, which passed, gave nuclear energy more opportunity for local tax exemptions.


This was good news and bad. There are two main motives behind the push for nuclear power. One is the desire to reduce global warming. The other is the interest in making money with data centers and AI.


Reducing global warming is needed. Some leaders believe increasing solar, wind and geothermal power to replace fossil fuels will not be enough. This may be. But before taking the leap to nuclear now, we need a thorough and independent study of the costs, benefits and risks of all the alternatives to coal and oil.


Nuclear power has been dangerous in the past, and there is no reason to assume it will not be dangerous in the future. We still struggle to manage the waste – “dispose” is not the word. And new nuclear is experimental, expensive, and will return no benefits for a long time. Meanwhile, many states are discovering the value of wind and solar power. The leader is Texas, where wind and solar now account for 30% of its large energy production.


In addition to reducing global warming by replacing fossil fuel, we do so by reducing demand for energy. An evaluation of the costs and benefits of alternatives to coal and oil should include this as well. Reducing demand, of course, is the complete opposite of the desire of data centers. Their demand for energy is increasing so fast that some states have reversed plans to reduce fossil fuel burning to meet their needs.


Data centers’ demands for rapid increases in power are straining electrical grids. The complexity and cost of this are leading to heavy debates on who should pay. Data centers and the AI tech behind them complain that they are being mistreated by this close attention. They are offering to build small nuclear reactors next to data centers as a friendly solution.


Let’s take a look. First, the idea that data centers need any public support is absurd. Tech is putting hundreds of billions of dollars into data centers, and promising more. They brag about it.


This year the governor discovered that data centers are creating few permanent jobs, so he proposed eliminating the tax exemption for their computer replacements. The legislature agreed and found

$60 million in ordinary taxes, a big help in a tough budget year. But if data centers don’t create permanent jobs, why are they still exempt from ordinary taxes on construction? Tech apparently prevailed.


The economic arguments presented to the legislature by data center advocates mainly described nice things they built for neighboring small towns. More and more, across the nation, small towns don’t like data centers. Voters in Festus, Missouri recently bounced all city council incumbents for approving a data center.


What is AI worth to us? It has many amazing abilities. Credibility is not one of them. Don’t ask AI to help with your tax return. Don’t believe a legal precedent claim made by AI. AI depends on plausibility, not validity, and increasingly is in its own spiral of following its own language.


AI needs good management. How is it being managed today? Follow the money. AI does what it is told to do. Currently it is told to seduce us into social media, maximize our time in it, and place ads customized to us. Big tech has begun to pay for the damage of this to young people, but they can afford it. Alongside this, tech employees who fear losing their jobs are competing by starting as many AI programs as they can. Performance and the massive load on data centers and energy consumption are irrelevant.


We might need limited and tightly controlled nuclear power in the near future to prevent more global warming. This needs careful and complete analysis, not driven by the Trump administration. In the meantime, do data centers need nuclear? Do we need data centers? Regulating data centers and the AI they serve is complicated. Giving nuclear power to big tech is not the answer.

Peter Harris is a member of PSARA's Climate and Environmental Justice Committee.

bottom of page